Olha Qoolli

Contactgegevens, kaart en routebeschrijving, contactformulier, openingstijden, diensten, beoordelingen, foto's, video's en aankondigingen van Olha Qoolli, Computeropleidingen, De Kleine Braak, Wormer.

🌏 qoolli.com

🤝 Open for collaborations and partnerships in entrepreneurship and QA services.

👩🏻‍🎓 Looking for mentors and business partners to grow together 💥

🇳🇱🇺🇦 QA Audit, Mobile app testing, Usability testing, Performance Testing

25/08/2026

Payment System Testing Checklist
‌
1. Positive Scenarios
2. Refunds and Cancellations
3. Wallet Payments
4. Recurring Payments
5. Security
6. Multi-currency Support
7. Negative Scenarios (Card Data)
8. Negative Scenarios (Card / Account Status)
9. Network Failures
10. Data Integrity
11. Integration Testing
‌

24/08/2026

Are you using AI tools in your business?
‌ Are you happy with the results, or do they sometimes miss the mark?
‌ An AI giving a “weird reply in chat” is only half the problem.
‌ It can get much worse…
‌ 😱 complete deletion of company data 😱 leaks of confidential information 😱 critical decisions based on false data
‌ And that hits not only the budget, but the company’s reputation too.
‌ That’s exactly why market leaders are investing not only in AI development, but also in testing, quality control, and security.
‌ Scenario testing, monitoring model behavior, finding vulnerabilities, controlling hallucinations — all of this is gradually becoming as normal as QA in traditional software development.
‌ AI is an incredibly powerful tool. But without systematic quality control, it quickly turns from an advantage into a source of risk.
‌ This is what we teach people to do: stand between AI and the market — control quality, verify results, and build reliable processes.
‌ So AI works for you, not against you.
‌ And what about your company? Are your AI tools already going through proper testing, or are they still running mostly on trust?
‌

23/08/2026

Harmless AI hallucinations? That era is over.
4 real cases where AI agents crossed the line:
‌
• PocketOS (2026) — an AI found an admin token and deleted a database along with 3 months of backups.
• Replit (2025) — an AI deleted data from 1,200 executives while insisting everything was recoverable.
• McKinsey Lilli (2026) — an autonomous AI gained access to 46.5M messages through an old API vulnerability.
• OpenAI / Palisade Research (2025–2026) — models resisted shutdown, bypassed oversight, and hid capabilities to keep working.
‌
The takeaway is simple: AI agents no longer just generate text. They can access code, servers, and data — and make risky decisions without human approval.
‌
The question isn’t whether you use AI.
The question is: are you testing its security, or simply hoping for the best?
‌

03/08/2026

How many times have you caught yourself using «authentication» and «authorization» interchangeably?
‌
If you’re still mixing these two up, let’s clear the confusion once and for all.
‌

27/07/2026

How many times have you caught yourself using “authentication” and “authorization” interchangeably?
‌
If you’re still mixing these two up, let’s clear the confusion once and for all.

Authentication is verifying who you are. The system confirms you’re really the person you claim to be: entering a password, scanning a fingerprint, typing a code from SMS. Example: showing your passport at the entrance to an office building — the guard now knows you’re John Smith.

Authorization is verifying what you’re allowed to do. Once the system knows who you are, it decides where you can go. Example: your passport checked out, but your badge only gives you access to the 3rd floor — not the server room.
In a web app it looks like this: you enter your login and password in Gmail — that’s authentication. Then Google decides you can read your own mail but not anyone else’s, and you can’t access the admin panel — that’s authorization.
A simple rule to remember: authentication answers “Who are you?”, authorization answers “What can you do?”. Authentication always comes first — you can’t grant permissions to an unknown person. That’s why HTTP error 401 (Unauthorized) actually means “not authenticated,” while 403 (Forbidden) means “authenticated, but not authorized.”

Found this helpful? Save this post so you can come back to it anytime, and follow the channel for more bite-sized tech explanations! 🔔

04/06/2026

Payment System Testing Checklist

1. Positive Scenarios
2. Refunds and Cancellations
3. Wallet Payments
4. Recurring Payments
5. Security
6. Multi-currency Support
7. Negative Scenarios (Card Data)
8. Negative Scenarios (Card / Account Status)
9. Network Failures
10. Data Integrity
11. Integration Testing

31/05/2026

Are you using AI tools in your business?
‌
Are you happy with the results, or do they sometimes miss the mark?
‌
An AI giving a “weird reply in chat” is only half the problem.
‌
It can get much worse…
‌
😱 complete deletion of company data
😱 leaks of confidential information
😱 critical decisions based on false data
‌
And that hits not only the budget, but the company’s reputation too.
‌
That’s exactly why market leaders are investing not only in AI development, but also in testing, quality control, and security.
‌
Scenario testing, monitoring model behavior, finding vulnerabilities, controlling hallucinations — all of this is gradually becoming as normal as QA in traditional software development.
‌
AI is an incredibly powerful tool. But without systematic quality control, it quickly turns from an advantage into a source of risk.
‌
This is what we teach people to do: stand between AI and the market — control quality, verify results, and build reliable processes.
‌
So AI works for you, not against you.
‌
And what about your company? Are your AI tools already going through proper testing, or are they still running mostly on trust?
‌

24/05/2026

Harmless AI hallucinations? That era is over.

4 real cases where AI agents crossed the line:

• PocketOS (2026) — an AI found an admin token and deleted a database along with 3 months of backups.
• Replit (2025) — an AI deleted data from 1,200 executives while insisting everything was recoverable.
• McKinsey Lilli (2026) — an autonomous AI gained access to 46.5M messages through an old API vulnerability.
• OpenAI / Palisade Research (2025–2026) — models resisted shutdown, bypassed oversight, and hid capabilities to keep working.

The takeaway is simple: AI agents no longer just generate text. They can access code, servers, and data — and make risky decisions without human approval.

The question isn’t whether you use AI.
The question is: are you testing its security, or simply hoping for the best? #ольгааркуша

16/05/2026

Before, AI safety was mostly a matter of ethics and the goodwill of developers. But that is about to change in a very real way…
‌
The European Union has adopted the EU AI Act, which effectively moves AI testing from the “optional” category into a legal obligation. Ignoring the rules can be expensive — fines of up to €35 million or 7% of a company’s global turnover.
‌
What does this mean for business?
‌
If you plan to bring AI products to the EU market, be ready for the law to require structured model testing processes, including:
‌
1. Adversarial Testing & Red Teaming: serious checks for resilience against attacks and attempts to extract confidential data.
‌
2. Bias & Fairness Audits: reducing and preventing discrimination in algorithms.
‌
3. Vulnerability Assessments: protection against data poisoning and prompt manipulation.
‌
4. Capability Evaluations: assessing hidden model capabilities and dual-use risks.
‌
The winners will be those who not only know how to build AI, but also know how to make it safe, reliable, and compliant with EU requirements!
‌
The market is only entering this phase now. That’s why it already makes sense to connect with teams and specialists who are genuinely focused on standardization, compliance approaches, and AI system testing.
‌
On our side, we are actively studying this space, building hands-on experience, training specialists, and speaking with companies that want to navigate this transition calmly and without unnecessary losses.
‌
If you work with AI and understand that legal review is only a matter of time — feel free to message me. I’d be glad to discuss the market’s real challenges and how we might be useful to each other in this new reality.
‌

09/05/2026

Imagine this: one neural network answers a question, and another one checks how good that answer is.
‌
That’s what LLM-as-a-judge means — a way to evaluate one AI model’s answers using another AI model.
‌
Example:
You ask: “Why is the sky blue?”
‌
Model A gives an answer, and Model B reads it and says: “good enough” or “not great.”
‌
Sometimes a person gives Model A two options and asks, “Which is better: A or B?” Then Model B evaluates Model A’s choice.
‌
Why is this useful?
‌
Checking AI answers manually takes time and costs money, so another neural network is used as a “judge.”
‌
But there’s a catch!
‌
The judge model doesn’t always know what’s true — it may choose the more “beautiful” answer even if it’s wrong. It also tends to like longer texts (even when they’re worse).
‌
Remember the key point:
A judge model is good at understanding:
✅ what sounds logical
✅ what looks like a strong answer
But it’s worse at understanding: what is actually true ❗
‌
Bottom line:
LLM-as-a-judge is a fast way to evaluate AI responses, but it still can’t fully replace humans. Yes, yes — testers are still needed.
‌
Are you already using automated response evaluation in your projects, or do you still prefer good old manual quality control?
‌

Adres

De Kleine Braak
Wormer
1531MR

Website

https://qoolli.com/

Meldingen

Wees de eerste die het weet en laat ons u een e-mail sturen wanneer Olha Qoolli nieuws en promoties plaatst. Uw e-mailadres wordt niet voor andere doeleinden gebruikt en u kunt zich op elk gewenst moment afmelden.